This paper introduces an outlier analysis which can improve the convergence of the statistical analysis results of sloshing model test data. The paper classify possible outliers in the sloshing model test into three categories and present a treatment method for each outlier. The developed outlier analysis is adapted to the model test results for the cargo of the liquefied-natural-gas (LNG) carrier in operation. The results of the present new method are compared with those of the conventional procedure, particularly focusing on long-term sloshing prediction. Through this study, the effectiveness of the present method is observed, and it is found that the present method provides is robust and reliable results in the application of experimental data for load prediction.